Abstract task sequence initiation deficit dissociates anxiety disorders from obsessive-compulsive disorder and healthy controls

Abstract

In everyday life, humans carry out sequences of tasks. These tasks may be disrupted in those with obsessive-compulsive disorder (OCD). Symptoms like compulsions can be considered sequential and often cause repetitions of tasks that disrupt daily living (e.g., checking the stove while cooking). Motor sequences have been used to study behavioral deficits in OCD. However, not all sequences are motor sequences. Some are more “abstract” in that they are composed of a series of tasks (e.g., chopping and stirring) rather than being dependent on individual actions or stimuli. These abstract task sequences require cognitive control mechanisms for their execution. Though theory has proposed deficits in these sequences in OCD as well, they have not been directly investigated. We tested the hypotheses that OCD participants exhibit deficits in the control mechanisms specific to abstract task sequences and more general flexible behavior (measured with task switching within the sequences), relative to health controls (HCs) and clinical controls (participants with anxiety disorders, ANX). One hundred twelve participants completed abstract task sequences consisting of simple categorization tasks. Surprisingly, participants with OCD did not perform worse than HCs or ANX. However, ANX participants showed impairments specific to sequential control that did not extend to more general flexible control. Thus, we showed a novel behavioral dissociation between OCD and ANX specific to abstract task sequential control. These results also implicate deficits in specific frontal sequential control neural circuitry in ANX and not in OCD, where implicit sequential deficits may more closely align with striatal circuits.

Background

Obsessive-compulsive disorder (OCD) is defined by repetitive and intrusive thoughts (obsessions) and actions (compulsions) which feel difficult to control and affects approximately 2-3% of the world population during their lifetime (Kessler et al., 2005; Lack, 2012; Veale & Roberts, 2014). Neurobiological models of OCD implicate cortico-striatal-thalamic-cortical (CSTC) circuitry (Graybiel & Rauch, 2000; Menzies et al., 2008; Milad & Rauch, 2012; Nakao, Okada, & Kanba, 2014; Shephard et al., 2021), but this relationship is multifaceted. For example, it has been suggested that cognitive process and related symptomology in OCD cannot be ascribed to singular components of the circuit, both because of ambiguities in the tasks used to assess deficits containing multiple potential processes and because of the involvement of networks of brain areas (Shephard et al., 2021). Therefore, utilizing multimodal tasks designed to dissect cognitive processes related to specific OCD symptomology, rather than that of closely related disorders such as anxiety disorders, may lead to more comprehensive models with which to dissociate behaviors and the putative neural circuitry involved.

Our goal was to study sequential behaviors potentially related to the compulsions that are specifically a part of OCD symptomology, and not of anxiety disorders, to better inform neurobiological models. Compulsions can be considered sequential in that they are repetitive mental or physical tasks or rituals (Lack, 2012). Our general approach was two-fold. First, we examined sequential behaviors that had a structure as parallel as possible to those in daily life in terms of the cognitive demands that extend through time (discussed further below). Second, we compared participants with OCD to both healthy controls (HC) and participants with anxiety disorders (ANX). ANX was the clinical comparison group because of its similarity to OCD in the experience of fear and anxiety (OCD was categorized as an anxiety disorder in previous editions of the Diagnostic and Statistical Manual of Mental Disorders, DSM). However, although excessive worry and rumination in ANX can produce repetitive behaviors, ANX does not result in the compulsions that are part of OCD (Goodwin, 2015). Further, it has been suggested that repetitive behaviors in ANX could be the response to the perception of threat (Goodwin, 2015) and the dysfunction of fear-related neural circuitry (Duval, Javanbakht, & Liberzon, 2015), whereas the compulsions in OCD may have different underlying neural circuitry (i.e. CSTC) (Harrison et al., 2009; Jung et al., 2013). In support of a putative relationship among compulsions, behavioral sequences, and neurobiological models of OCD, deficits have been previously observed in OCD in a specific sequence paradigm: implicit motor sequences. Implicit motor sequences are sequences of motor actions (e.g., button presses) that are learned by repeating the same actions over time without explicit instruction as to the existence of the overarching sequence. For example, a motor sequence that could be implicit is playing a scale on a piano or other instrument. The player is cued by a series of notes that result in specific motor actions in the required order. The musical sequence (e.g., scale or melody) changes if any key press actions or ordering changes. The performance of implicit motor sequences is striatal-dependent (Janacsek et al., 2020), which is a component of the CTSC circuitry implicated in OCD. Thus, a behavioral deficit in implicit motor sequences would support a potential dysfunction in CTSC circuitry in OCD that is more reliant on striatal activity and potentially inform the use of treatments that selectively target these regions. Indeed, several studies have shown deficits in implicit motor sequences (in the form of increased reaction times, RTs) in OCD compared to healthy controls (HCs) (Kathmann, Rupertseder, Hauke, & Zaudig, 2005; Kelmendi et al., 2016; Soref, Liberman, Abramovitch, & Dar, 2018) and compared to anxiety disorders (social anxiety, panic disorder, and agoraphobia) (Goldman et al., 2008). This dissociation between OCD and anxiety disorders supports the idea that behavioral sequence paradigms can be useful as a window specifically into OCD symptomology.

However, there are limitations to studying sequential behavior using implicit motor sequences in OCD. First, ambiguity exists as to the specific elements of implicit motor sequences that may be impaired in OCD. For example, a common version of an implicit motor sequence task is the serial reaction time (SRT) task. In the SRT task, a visual cue indicates the location of the button that the participant should press on each trial. Unbeknownst to the participant, the cued button presses are part of a repeating sequence. Deficits on this implicit motor sequence task could be due to an impairment in the ability to learn the sequence or an impairment in the ability to perform the visuomotor responses (Brezóczki et al., 2023; Soref et al., 2018). Second, there are other kinds of sequences such as explicit (instructed) sequences and non-motor sequences. It is unknown if the deficits exhibited in OCD on implicit motor sequences is specific, or if it extends to other kinds of sequences as well (Brezóczki et al., 2023; Soref et al., 2018). Therefore, implicit motor sequences may not be a complete description of sequential behaviors that relate to OCD.

Other models of sequential processing in OCD encompass sequences that are not necessarily implicit motor sequences (Huey et al., 2008), but this model has not been directly tested. This model emphasizes the role of the prefrontal cortex (PFC) in memories of behavioral sequences they define as “structural event complexes”. This broadening of the conceptualization of “sequence” in OCD is complemented by the observation that many of the repetitive ordering and mental compulsions in OCD cannot be classified as implicit motor sequences. For example, repeatedly arranging objects in a specific order (Lin & Gao, 2022) or counting objects in sets of five (Menon, 2013). Although these sequences require actions, they are not dependent on the precise identity or location of the objects being arranged or counted. These sequences depend on a rule that dictates how they should be carried out (e.g., the counting of objects must be in sets of five for the sequence to be complete). To illustrate this distinction by example, rearranging the piano keys would be disastrous for an implicit motor sequence (because it changes the motor actions), whereas in these compulsion-related sequences rearranging/counting writing implements would proceed regardless of if they were pens or pencils, scattered or piled on a table (because it’s not dependent on the specific arm/hand motion). Long recognized as a key component of daily behavior, we define these sequences as abstract task sequences (Desrochers et al., 2022; Lashley, 1951). Abstract task sequences contain an ordered set of operations through time with a beginning and end, but, importantly, they are not bound by exact action, time, or stimuli. For example, when cooking pasta, the process proceeds whether the tomato is retrieved from the counter, the garden, or the refrigerator or whether the slicing is performed on the tomato or onion. That is, abstract task sequences isolate sequential components from visuomotor associations (that can create ambiguity in implicit motor sequences). There is evidence that abstract task sequences depend on rostral PFC (Desrochers, Chatham, & Badre, 2015; Desrochers, Collins, & Badre, 2019). Rostral PFC is interconnected with CSTC circuits yet distinct from striatal-dependent implicit motor sequences. To our knowledge, abstract task sequence performance and its theorized reliance on the PFC has not been empirically tested in OCD. Understanding abstract task sequences in OCD will expand currently limited knowledge of the nature of sequential behaviors in this clinical population and potentially more closely approximate complex sequential compulsions. If our hypotheses are supported, results could suggest an additional region of the PFC (rostrolateral PFC), which is heavily recruited during this cognitive process, be included in CSTC models of OCD.

To test abstract task sequences and their relationship to OCD, it is first necessary to define their components. These sequences can be thought of as hierarchical, with the overarching sequential goal (e.g., make pasta) dictating the order of sub-tasks (e.g., chopping) (Desrochers et al., 2019; Schneider & Logan, 2006). The associated cognitive processes to accomplish this ordering of tasks are referred to as sequential control and cognitive control, respectively. Importantly, these processes are separable both behaviorally (Schneider & Logan, 2006; Trach, McKim, & Desrochers, 2021) and neurally (Desrochers et al., 2022, 2015, 2019) in healthy populations. Whether sequential control is impaired in OCD in a manner similar to deficits previously observed in implicit motor sequences remains unknown. In contrast, cognitive control (not in a sequential context) has been studied in OCD. Two common measures are relevant to investigating cognitive control: task switching and post-error slowing. Both, discussed below, are measures of the flexibility that is likely necessary for cognitive control that is a constituent of abstract task sequences.

Studies of task switching inform our understanding of flexible cognitive control in OCD. Task switching (e.g., transitioning from chopping to stirring) is a specific facet of cognitive control that is important for executing abstract task sequences, cognitive flexibility. Specifically, task switching is flexibly moving from one set of rules to another (Monsell, 2003), and can be considered an adaptive behavior in executing abstract task sequences successfully. Task switching deficits have been observed in OCD compared to HCs (Gu et al., 2008; Han et al., 2011; Liu et al., 2023; Meiran, Diamond, Toder, & Nemets, 2011; Remijnse et al., 2013). One early study did fail to observe this deficit (Moritz, Hübner, & Kluwe, 2004), though it has been hypothesized that this result may have been due to co-morbid depression or medication status (Remijnse et al., 2013). At least one study observed a positive correlation between OCD symptom severity and neural activity related to task switching deficits (Remijnse et al., 2013) and another observed reduced deficits after OCD treatment (Han et al., 2011), further strengthening the connection between task switching as a measure of flexible cognitive control and OCD.

Post-error slowing is another measure of flexible cognitive control potentially important to abstract task sequence execution, and non-sequential studies of this effect in OCD provide a foundation. Post-error slowing is a delay in reaction time on a trial following a previous error and is thought to be an adaptive behavior by reflecting the capacity to flexibly adjust behavior based on a previous outcome (Dutilh, van Ravenzwaaij, et al., 2012). This kind of flexibility may also be important when completing task sequences (e.g., slow slicing after a wrong cut). Post-error slowing has not often been studied in the context of OCD and the results are mixed, possibly because the phenomenon can be studied in the context of nearly any task and the specific task itself may play a role. One study used a flanker task with congruent and incongruent stimuli and showed deficits in post-error slowing in participants with OCD compared to HCs (Modirrousta, Meek, Sareen, & Enns, 2015), while another that used a stop-signal task did not observe a difference in participants with OCD, but did in patients with ANX (Rueppel, Mannella, Fitzgerald, & Schroder, 2022). Related to symptomatology, excessive post-error slowing could reflect maladaptive behavior in OCD, such as rumination on past errors or worry about future errors. Therefore, an open question remains as to whether and how post-error slowing is affected in OCD, and the fact that it can be measured in the context of a variety of tasks can be an advantage.

Our goal was to bring together these elements (sequential and flexible cognitive control) to comprehensively test if behavioral sequences that go beyond implicit motor sequences are impaired in OCD. Doing so would determine if sequence deficits in OCD (observed in implicit motor sequences) extend beyond visuomotor processes to sequential and flexible cognitive control. The results of these experiments would have implications for neurobiological models of OCD as to which parts of CSTC circuitry (e.g., frontal or striatal) may be implicated in sequential deficits and perhaps in the disorder as a whole. This knowledge may direct more precise treatments for OCD targeting a specific part of the CSTC as a function of an individual’s degree of sequential impairment.

To accomplish this goal, we used a previously developed abstract task sequence paradigm that integrated task switching in sequences to dissociate specific sequential control from more general flexible control (Desrochers et al., 2022, 2015, 2019; Schneider & Logan, 2006; Trach et al., 2021). Though the stimuli and tasks are relatively simple, the structure more closely approximates the multiple control mechanisms (and other cognitive resources) extended through time in daily life than implicit motor sequences. We compared the performance of participants with OCD to two groups: participants with anxiety who do not have OCD (ANX) and healthy controls (HC). These groups mirror previous studies of implicit motor sequences and provide a clinical control. After determining if sequential and cognitive control are separable in each group (using sequence initiation and task switching, respectively), we tested two main hypotheses. First, we hypothesized that participants with OCD compared to HC and ANX would show sequential control deficits based on previous findings in implicit motor sequences and observations that some compulsions in OCD can be conceptualized as abstract task sequences. Further, we hypothesized that these deficits would not be explained by other measures of flexible control, such as post-error slowing, and would be positively associated with symptom severity. Second, we hypothesized that as in non-sequential task-switching, we would observe deficits in task switching within sequences for participants with OCD relative to ANX and HC that are separable from measures of sequential control, as in non-sequential task switching, and that these task switching deficits would also positively correlate with symptom severity.

Methods

Experimental Procedures

Participants

Participants were part of a larger study investigating the neural bases of putative dimensional endophenotypes (harm avoidance and incompleteness) underlying symptoms of obsessive-compulsive related and anxiety disorders. In the larger study, participants underwent clinical screening and completed several cognitive tasks. Participants were recruited in community and clinical settings in and around Providence, RI. Cognitive tasks and interviews were administered by trained evaluators (see description of, below). All participants gave informed, written consent approved by the Institutional Review Board.

Clinical participant inclusion criteria for the larger study were as follows: Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) primary diagnosis of the following obsessive-compulsive related or anxiety disorders: OCD, obsessive-compulsive personality disorder (OCPD), hoarding disorder, panic disorder, agoraphobia, or social anxiety disorder, 2) age 18 to 65, 3) English speaking, and 4) willing and able to provide written informed consent. Clinical exclusion criteria: 1) Cognitive impairment (organic brain syndrome, dementia) that would interfere with study participation, ability to provide informed consent, or completion of self-report questionnaires, 2) current psychotic disorder, 3) psychiatric medications other than serotonin reuptake inhibitors (SRIs) or medications taken for sleep or occasional anxiety (e.g., hydroxyzine, trazodone, etc.), 4) pre-morbid IQ < 85 as measured by the National Adult Reading Test (NART), 5) Implanted metallic substances, metallic tattoos received prior to 1990; and 6) pregnancy and any other conditions not allowed in the scanner that would represent a safety risk for participants.

Inclusion and exclusion criteria for the healthy controls were the same as for clinical participants except for no current (past month) DSM-5 diagnosis of any psychiatric disorder or a lifetime diagnosis of OCD or related target disorder (OCPD, hoarding disorder), any anxiety disorder, psychotic disorder, or bipolar mood disorder.

Data from one hundred thirty-nine (105 female) adults (ages 18-64 years; mean 33.6 years) were initially included the current study. Fifteen participants were removed from subsequent data analysis due to not completing the abstract sequence behavioral task. From the remaining 124 participants, 12 were excluded from further data analysis for high error rates (ERs) (> 20% overall) (Desrochers et al., 2015, 2019; Schneider & Logan, 2006; Trach et al., 2021). As in previous experiments using this task, error rates above 20% indicate these participants may not have followed instructions, and thus may not have been fully engaged in the task. This resulted in a total of 112 (86 female) participants (ages 18-64; mean 32.4 years, standard deviation 12.4 years) that were included in data analysis. Across all included participants, the average ER = 5.05%, +/− 4.19 SD, with a median of 3.43% and range of 0.2 – 17.06 %.

Here, we report data from participants who completed the abstract task sequential control paradigm and had a primary diagnosis of either OCD (OCD, n=43), a primary diagnosis of an anxiety disorder (panic disorder, agoraphobia or social anxiety disorder; ANX, n=22) without comorbid OCD, or were a healthy control (HC, n=47). Only anxiety disorders were included in the ANX group. Clinical participants with a primary diagnosis included in the larger study that was not included in the OCD or ANX group, i.e., those with primary OCPD or hoarding disorder, were excluded from analysis in this study. Within the primary OCD group, 32.6% (n=14) of individuals had primary OCD and no other diagnoses of disorders that were assessed, 13.9% (n=6) of individuals had primary OCD and a comorbid (non-primary) anxiety disorder (panic disorder, agoraphobia, social anxiety disorder), and 53.5% (n=23) of individuals had primary OCD and a comorbid (non-primary) related disorder that was assessed (OCPD, hoarding disorder, body dysmorphic disorder, post-traumatic stress disorder). Table 1 summarizes the demographics for the HC, OCD, and ANX groups.

Table 1. Group descriptive statistics.

Descriptive statistics for the three groups of interest (OCD, ANX, HC). Information provided includes number of group members, average age (+/− 1 standard deviation), number of males and females, average Y-BOCS (range 0-40), average total APPQ (range 0-216), and race/ethnicity. See methods for details about co-morbidities in the OCD and ANX groups.

Demographics	OCD	ANX	HC
No. subjects	43	22	47
Age, years, mean (S.D.)	31.1 (12.7)	36.3 (12.8)	31.2 (11.8)
Sex, number F(M)	35 (8)	12 (5)	34 (13)
Y-BOCS total, mean (S.D.)	21.4 (5.0)	5.8 (7.1)	0.4 (2.3)
APPQ total, mean (S.D.)	58.1 (34.5)	54.5 (29.2)	7.7 (10.8)
Race/ethnicity (% Caucasian/% Hispanic)	74.4/21	77.3/13.6	55.3/10.6

Measures

Anxiety and Related Disorders Interview Schedule for DSM-5 (ADIS-5) (Brown & Barlow, 2014):

The ADIS-5 is a semi-structured interview designed to determine reliable diagnosis of DSM-5 disorders and to screen for the presence of other conditions. Here, the ADIS-5 was used to assess diagnosis of the following DSM-5 disorders: OCD, panic disorder, social anxiety disorder, agoraphobia, hoarding disorder, body dysmorphic disorder, and post-traumatic stress disorder, and to provide ratings of symptom severity. Primary diagnosis was determined by the disorder with the highest clinical severity rating (CSR).

Structured Clinical Interview for DSM-5 Personality Disorders (SCID-5-PD)- OCPD module (First, Williams, Karg, & Benjamin, 2016):

The SCID-PD is a clinician administered semi-structured interview designed to assess personality disorders. Here, the SCID-5 OCPD module was used to assess for OCPD.

Yale-Brown Obsessive Compulsive Scale (Y-BOCS) (Goodman et al., 1989):

The Y-BOCS is a gold-standard rater-administered interview that assesses the presence and severity of obsessions and compulsions over the past week.

Albany Panic and Phobia Questionnaire (APPQ) (Rapee, Craske, & Barlow, 1994):

The APPQ is a 27-item measure of the dimension of fear activities that produce physical sensations (e.g., exercise) and fear of common agoraphobia and social phobic situations. The measure has three subscales, interpreted as reflecting fear of agoraphobic situations (“Agoraphobia”, 9 items), fear of activities that produce somatic sensations (“Interoceptive”, 8 items), and fear of social situations (“Social Phobia”, 10 items). Internal consistency (measured using Cronbach’s Alpha) in the OCD group for “Agoraphobia”, “Interoceptive”, and “Social Phobia” subscales were 0.84, 0.81, and 0.78 and for the ANX group were 0.92, 0.69, and 0.81, respectively.

Procedure

The behavioral task used was the same as in a previous neuroimaging study of HCs (Desrochers et al., 2015) (Fig. 1) and was based on previous behavioral tasks used to study abstract task sequential control (Schneider & Logan, 2006). Participants were presented on each trial with a stimulus of varying size (small [3.5 x 3.5 cm] or large [7 x 7 cm]), shape (circle or square), and color (blue or red) (Fig. 1A). The combination of stimulus size, shape, and color formed 8 possible stimuli, which appeared equally throughout the task and did not repeat on adjacent trials. A white fixation cross was displayed during each intertrial interval (ITI) after each stimulus, which was 0.5 s throughout the task. Each trial was shown with response options for the color and shape of the stimulus, mapped onto ‘j’ and ‘k’ keyboard keys. Trials timed out after 4 s if no key response was made. Responses were mapped from two fingers, the index and middle of the dominant right hand, onto the ‘j’ and ‘k’ keys. Each key corresponded to one shape and color combination (e.g., ‘j’ maps to both ‘blue’ and ‘circle’ while ‘k’ maps to both ‘red’ and ‘square’). Participants pressed one button per trial to indicate their choice of color or shape. These stimulus-response mappings were kept the same throughout the task for each participant and were counterbalanced across participants. The frequency of responses to each stimulus and repeat button presses were counterbalanced throughout the task.

Figure 1. Behavioral task schematic.

A) Example trials in a block for the simple sequence. Each block begins with a screen that instructs the sequence, e.g., “COLOR, COLOR, SHAPE, SHAPE”. Each trial consists of one stimulus presentation where the participant must make the correct categorization decision based on the identity of the stimulus and the position in the sequence. The remembered categorization decision for each item is indicated in a thought bubble and the correct choices for each trial are indicated by black arrows. The stimulus remains on screen until a response is made (max 4 sec). After the response (or response time-out), a fixation cross is displayed for the duration of the intertrial interval (ITI, 500 ms). Distance between images is for illustration purposes only and does not represent actual timing. There are approximately 24 trials per block, it can end on any position in the sequence, and the block ends with a sequence position question asking, “What is the NEXT item in the sequence?”. B) Example run containing four blocks, with each block being a simple (CCSS [color, color, shape, shape]; SSCC [shape, shape, color, color]) or complex (CSSC [color, shape, shape, color]; SCCS [shape, color, color, shape]) sequence. The order of the blocks is counterbalanced across the five runs that each participant performs.

Stimuli were presented in blocks (Fig. 1B) (24-27 trials each, so that blocks ended on different and unpredictable sequence positions, counterbalanced across blocks), with participants completing 4 blocks per run for a total of 5 runs. At each block start, a 4-item sequence was displayed (5 s), followed by a fixation screen (1 s). The items in the sequence (e.g., color, color, shape, shape) indicated the choice a participant should make for each stimulus. In this example, the choice for the first trial corresponds to the image color, for the second trial the image color, for the third the image shape, and for the fourth the image shape. Participants had to remember the sequence throughout the block and re-initiate every 4 stimuli until the end of the block. No cues were given to participants to indicate sequence position throughout the block. Participants did not receive feedback, in any form, throughout the task.

At the end of each block, a question was displayed which asked participants to choose which sequence item would occur next if another stimulus appeared. Participants responded with one of four keys (‘j’, ‘k’, ‘l’, and ‘;’) to indicate which sequence position (1, 2, 3, or 4) would have come next. These trials timed out after 10 s if no response was recorded. After this screen, a fixation cross was displayed, followed by instructions for the next block.

Each block consisted of one sequence type, of which there were two total (Fig. 1B). Simple sequences followed the pattern AABB (with ‘A’ corresponding to one choice type and ‘B’ corresponding to the other, i.e., AABB corresponds to the sequences color, color, shape, shape and shape, shape, color, color) and are termed ‘simple’ for having one task-switch (from A to B) within the sequence. Complex sequences followed the pattern ABBA (i.e., shape, color, color, shape and color, shape, shape, color) and are termed ‘complex’ for containing two task-switches (from A to B and B to A) within the sequence (Schneider & Logan, 2006). Although the number of switches differed within each sequence, the number of task switches was equivalent across sequences throughout the block as participants repeated sequences, making the probability of occurring switch or repeat trials equal between blocks of complex and simple sequences (i.e., the first position of simple sequences is a task switch). Therefore, inclusion of two different sequence types controls for task switching effects, particularly at the first sequence position.

Each run consisted of each sequence possibility (i.e., the two possible simple and two possible complex sequences), making a total of 4 blocks per run (Fig. 1B). The order of sequence blocks was counterbalanced across the runs.

Participants were trained on the task with 4 practice sequences prior to completing the experiment. Training first included button press practice for color and shape choices individually. Participants were then directed by the experimenter through one sequence practice, and finally practiced three more sequences without experimenter guidance. Once performance competency was established through training, participants began the experiment.

Behavioral Analysis

As in previous studies using this or related tasks (Desrochers et al., 2015, 2019; Schneider & Logan, 2006; Trach et al., 2021), the following sets of trials were excluded from analyses. The first sequence (four trials) of each block was excluded from analysis (approximately 3.5% of trials per participant) to prevent changes in reaction times (RTs) due to initiating a block from confounding changes in RTs due to initiating sequences or switching tasks. Additionally, trials with RTs < 100 ms were excluded (< 1% of trials per participant) because the knowledge of the correct button press could not be determined until the categorization decision was made, and thus short RTs could only be possible by chance. Error rates (ERs) were calculated for the remaining trials. Trials were also excluded on which participants “lost track” of the sequence position. To determine these trials, periods of 2 or more error trials were monitored and marked as “lost” periods until the next 4 correct adjacent trials occurred (approximately 3% of trials per participant). These trials were excluded to ensure that all data analyzed were trials in which we were certain participants were performing the task as instructed. Thus, the remaining errors used for analysis were not due to more globally losing track of the abstract task sequence. It is noted that because of the design of the responses in the task where one button is used to indicate both one of the colors and one of the shapes (e.g., one button indicates both “red” and “circle”) it is not possible to distinguish errors due to the correct task and wrong choice (e.g., choosing “blue” instead of “red” with a red square for a stimulus on a “Color” trial) from having the wrong task and correct choice (e.g., choosing “circle” instead of “square” with a red square for a stimulus on a “Color” trial) from a single trial alone. This design was chosen to replicate previous, established studies (Desrochers et al., 2015; Schneider & Logan, 2006).

Statistical analyses of performance were assessed using repeated measures analysis of variance (rmANOVA). Age was included as a covariate in all ANOVAs to acknowledge the potential impact age has on cognitive tasks (Artuso, Cavallini, Bottiroli, & Palladino, 2017) and due to a larger age range in the present study (18 – 65) relative to previous studies of abstract sequential control (18 – 35) (Desrochers et al., 2015, 2019; Trach et al., 2021). Sequence initiation was measured by initiation cost, calculated by subtracting RTs at position 3 from position 1 in the sequence (Desrochers et al., 2015; Schneider & Logan, 2006). Positions 1 and 3 were chosen for the initiation cost measure since they were always matched as either task switches or repeats in the sequence (i.e., in each sequence type, positions 1 and 3 are always either both repeats of or switches from the previous trial). Initiation costs were correlated with symptom severity measure composite scores and subscale scores, when applicable.

Bayes factor (BF₁₀) analyses were conducted on rmANOVAs and t-tests to provide further evidence for the null hypothesis in group comparisons when applicable. A BF₁₀ < 0.1 indicates strong evidence for the null hypothesis, while a BF₁₀ between 0.1 – 0.33 provides moderate evidence for the null hypothesis, and a BF₁₀ between 0.33 – 1 provides anecdotal evidence for the null hypothesis (Andraszewicz et al., 2014).

Inverse efficiency scores (IES) were calculated to investigate speed-accuracy trade-offs between OCD, ANX, and HC across each sequence position. IES combines RTs and ERs (Townsend & Ashby, 1978). We calculated IES by first obtaining a mean RT and ER at each sequence position for each participant. IES at each position was then calculated by dividing the mean RT by (1 – mean ER).

Many participants in the OCD and ANX groups had one or more co-morbidities. To assess if co-morbid disorders in these groups could account for additional variance in our behavioral data, we repeated rmANOVA analyses including a categorical covariate for participant subgroup. Participant subgroup was determined by the unique combination of comorbid diagnoses, e.g., OCD + hoarding disorder was labeled subgroup “A” and OCD + social anxiety + hoarding disorder was labeled subgroup “B”. Subgroups were determined separately for the OCD and ANX groups. All analyses were conducted using Matlab 2017b.

Since this study was part of a larger study that was optimized for different research goals, we were only able to conduct a post-hoc power analysis (Erdfelder, Faul, & Buchner, 1996) after the sample was collected. Given our sample size total, we were powered to detect a small effect size of 0.23 (η_p²) (Cohen, 1988) with an alpha of 0.05 and 80% power.

Results

To investigate sequential and flexible cognitive control, participants in three groups: OCD, ANX (anxiety disorder without OCD), and HC, completed five runs (each containing four blocks) of abstract task sequences for this experiment (Figure 1, see Table 1 and Methods for group composition). At the start of each block, participants were shown a four-item sequence, which they used to inform decisions about the stimulus color and shape on each trial (Figure 1A). Sequences followed the pattern ‘AABB’ (simple, containing one task switch, e.g.: color, color, shape, shape) or ‘ABBA’ (complex, containing two task switches, e.g.: color, shape, shape, color) (Figure 1B). Only one sequence was performed during each of the four blocks during a run (order counterbalanced across runs). Participants were not provided external sequence cues and had to internally keep track of the sequence throughout each block to respond correctly to each trial. Importantly, participants could not anticipate the motor response (button press) on each trial because it was contingent both on the identity of the stimulus (pseudo-randomly presented) and the position in the sequence. Overall, the participants across these groups performed the task as instructed and well (average error rate [ER] = 7.78, +/− 8.49 [1 SD]), with no difference in overall ER across groups (one-way ANOVA: F(2,111) = 0.02, p = 0.98).

We first tested the OCD, ANX, and HC groups separately for behavioral indicators of sequential and flexible cognitive control that had previously been observed in HCs (Desrochers et al., 2015, 2019; Schneider & Logan, 2006). We examined abstract sequential control using initiation cost as an indicator. Initiation cost is the increase in reaction time (RT) at the start of the sequence, relative to subsequent positions in the sequence. In this task, we measured initiation effects over and above any potential effects of switching or repeating tasks and therefore calculated it as the difference in RTs between positions in the sequence where task switching or repeating was “matched”. For both sequences, position three matches the task type at position one (e.g., in the complex sequence, ABBA, positions one and three are both task repeats). We replicated previous results in HCs and found that all groups exhibited significant initiation costs (Figure 2A; Table 2). As a control, in the OCD and ANX groups we also performed these analyses with a categorical covariate for comorbidity sub-group. We did not find any significant effects of this covariate (OCD: F(16,26) = 1.74, p = 0.1; ANX: F(2,19) = 0.93, p = 0.41). Thus, all participant groups performed the abstract tasks as sequences and showed evidence of sequential control.

Figure 2. General behavior results for OCD, ANX, and HC groups.

We report RT results in the first row and ER results in the second row. A) ANX exhibit significantly increased RT at position 1 compared to positions 2,3,4 compared to the HC group and OCD groups, and significantly higher initiation costs compared to the HC group. ANX exhibits significantly higher RTs across all positions compared to the HC and OCD groups. B) RT switch costs are not significantly different between any groups. (C,D) ERs across position and ER switch costs are not significantly different between any groups.

Table 2. Within groups rmANOVA for initiation cost.

Repeated measures ANOVAs were conducted on each individual group for position 1 minus 3 RTs (s). Mean initiation costs (+/− 1 S.D.), F statistics (d.f.s), P values, and effect sizes (η_p²) are reported in each column.

Group	Mean (S.D.)	F Statistic (d.f.s)	P	η p 2
OCD	0.31 (0.15)	171.02 (42)	<0.001	0.81
ANX	0.36 (0.15)	126.71 (21)	<0.001	0.86
HC	0.26 (0.14)	161.61 (46)	<0.001	0.78

We examined flexible cognitive control using switch costs as a measure of task switching. Switch cost is the difference between switching and repeating tasks (e.g., in AABB, the difference between positions two and three). For this analysis, we examined both ERs and RTs, as switch costs have been observed in both measures in HCs (Desrochers et al., 2015, 2019; Schneider & Logan, 2006). We replicated those previous results and observed switch costs in all groups independently (Table 3; Figure 2B, D). As in the RT analyses, we also performed a control analysis including a categorical covariate for comorbidity sub-group in OCD and ANX and again found no significant effects across the switch cost analyses (OCD: F(16,26) = 0.69, p = 0.78; ANX: F(2,19) = 0.26, p = 0.77). Therefore, all groups showed evidence of consistent switch costs and flexible cognitive control.

Table 3. Within groups rmANOVA for RT (s) and ER (%) switch cost.

Repeated measures ANOVAs were conducted on each individual group for switches – repeats in both reaction times (s) and error rates (%). Mean switch costs (+/− 1 S.D.), F statistics (d.f.s), P values, and effect sizes (η_p²) are reported in each column. Switch costs were calculated for each participant by subtracting the mean RT and ER across all repeat trials (pooled complex position 3, simple positions 2 and 4) from the mean RT and ER across all switch trials (pooled complex positions 2 and 4, simple position 3).

	Reaction time				Error rate
Group	Mean (S.D.)	F Statistic (d.f.s)	P	η p 2	Mean (S.D.)	F Statistic (d.f.s)	P	η p 2
OCD	0.18 (0.10)	142.77 (42)	<0.001	0.77	2.63 (3.41)	25.76 (42)	<0.001	0.38
ANX	0.19 (0.15)	33.11 (21)	<0.001	0.61	1.81 (3.62)	11.89 (21)	0.0031	0.37
HC	0.19 (0.15)	78.01 (46)	<0.001	0.63	2.67 (3.23)	31.01 (46)	<0.001	0.41

Having now established that markers of sequential and flexible cognitive control were present across the participant groups independently, we tested the first hypothesis that participants with OCD will exhibit abstract sequence performance deficits compared to HC and ANX groups, similar to findings in previous studies of implicit motor sequences (Goldman et al., 2008; Kathmann et al., 2005; Kelmendi et al., 2016). To test this hypothesis, we performed planned comparisons between initiation costs in OCD to HC and OCD to ANX groups. When comparing OCD to HC, we did not find a difference in RT across all positions in the sequence or initiation costs (Table 4; Figure 2A). Using Bayes factor analyses, we found moderate evidence for the null hypotheses in initiation cost (BF₁₀ = 0.16). We found a marginal difference between the OCD and ANX groups; however, it was not in the hypothesized direction. We found significantly greater RTs across all sequence positions and marginally greater initiation costs in the ANX group compared to the OCD group (Table 4; Figure 2A). Follow up testing also revealed that participants with ANX exhibited significantly higher RTs and initiation costs compared to the HC group (Table 4). Though we did not have a priori hypotheses about initiation cost in ER because it has not been observed in previous, similar studies (Desrochers et al., 2015, 2019; Schneider & Logan, 2006; Trach et al., 2021), for completeness we also compared ERs at positions 1 and 3 between groups. The OCD group had slightly lower ERs at position 1 that produced marginal interactions between OCD and HC (group x position: F(2,87) = 3.8, p = 0.054, η_p² = 0.042) and between OCD and ANX (group x position: F(2,62) = 1.9, p = 0.17, η_p² = 0.03). In summary, findings did not support the first hypothesis that OCD has greater initiation costs than HC and ANX and instead found evidence suggesting that ANX participants may have initiation cost deficits compared to OCD and HC.

Table 4. Between groups rmANOVA for initiation cost.

Repeated measures ANOVAs were conducted between groups for initiation cost (RT (s) position 1 – RT position 3). Statistics are reported for group, position, and group x position effects (factors). F statistics (d.f.s), P values, and effect sizes (η_p²) are reported for each factor.

Group	Factor	F Statistic (d.f.s)	P	η p 2
OCD vs. HC	Group	0.62 (87)	0.43	0.01
	Position	53.28 (87)	<0.001	0.38
	Group x Position	1.95 (2,87)	0.17	0.02
OCD vs. ANX	Group	4.79 (62)	0.03	0.07
	Position	60.96 (62)	<0.001	0.49
	Group x Position	3.25 (2,62)	0.08	0.05
ANX vs. HC	Group	6.68 (66)	0.01	0.09
	Position	37.56 (66)	<0.001	0.36
	Group x Position	7.61 (2,66)	0.01	0.11

Originally, a component of our first hypothesis was that a sequence initiation deficit in OCD would not be better explained by lower-level cognitive control mechanisms such as a speed-accuracy tradeoff or post-error slowing. However, we did not observe a sequence initiation deficit in OCD (compared to HC or ANX), and we did observe a sequence initiation deficit in ANX. Therefore, as a post-hoc follow-up to these results, we examined whether other, non-sequential measures of flexible cognitive control could better explain the observed initiation deficit in ANX. First, it was possible that, for participants with ANX, slower responses enabled decreased ER at the first position in the sequence (i.e., a speed-accuracy tradeoff). If this were the case, there would be a negative correlation between RTs and ER (such that increases in RT would lead to decreases in ER). We did not find evidence of this negative correlation in the ANX group, but rather a marginally positive correlation (p = 0.079, r = 0.38, slope = 5.97) at the first position (Figure 3A). A positive correlation between RT and ER is commonly observed during cognitive tasks (C. C. Wood & Jennings, 1976). Further, ERs at position 1 between groups were not significantly different (independent samples t-test, ANX vs. HC: t(67) = −0.089, p = 0.93; ANX vs. OCD: t(66) = 0.39, p = 0.69). There was moderate evidence for the null hypothesis between groups in ERs at position 1 (ANX vs. HC: BF₁₀ = 0.17; ANX vs. OCD: BF₁₀ = 0.18). To provide further evidence of a lack of speed-accuracy trade-off between groups, we used RTs and ERs to calculate inverse efficiency scores (IES) and compared the groups at position 1 (see Methods for details). We did not find a significant difference in IES among the three groups at position 1 (between-groups rmANOVA: F(2,109) = 0.022, p = 0.98). These results indicate that a speed-accuracy trade-off did not underlie the observed sequence initiation difference in ANX, supporting the hypothesis that sequential control deficits could not be better explained by more general mechanisms of flexible control.

Figure 3. Cognitive control measures do not lead to a sequence initiation difference in ANX.

A) Correlations between RT and ER do not indicate a speed/accuracy tradeoff at position 1 in the ANX group. B) Post-error RTs at position 1 in the ANX group are not significantly different than those in the HC group.

Another possible behavior that could account for significantly higher initiation costs in the ANX group is post error slowing. Post error slowing is an increase in RT on the response following an incorrect response and can be used as a marker of cognitive control, as this process is thought to reflect the capacity to flexibly adjust behavior after an outcome (Dutilh, van Ravenzwaaij, et al., 2012; Dutilh, Vandekerckhove, et al., 2012). In non-sequential tasks, increased post error slowing (greater RTs following error trials) has been observed in ANX compared to both OCD and HCs (Rueppel et al., 2022). In the current experiment, increased post error slowing in response to errors at position 4 could give rise to increased RTs at position 1 and the appearance of an increased initiation cost in participants with ANX compared to HCs. This post error slowing effect could therefore account, at least in part, for the increased initiation costs observed in ANX (Figure 2). To test this possibility, we first compared ER at position 4 between groups and then RTs at position 1 following position 4 errors between groups. We found that there was significant overall post error slowing in each group separately (within-subjects rmANOVA; HC: F(1,44) = 110, p = 1.46e-13, η_p² = 0.73; ANX: F(1,21) = 92.6, p = 3.78e-09, η_p² = 0.82; OCD: F(1,45) = 88.9, p = 3.20e-12, η_p² = 0.66). We did not find that the ANX group made significantly more errors at position 4 compared to the HC group (independent samples t-test, t(67) = −0.29, p = 0.78; Figure 2B). This result contained moderate evidence for the null hypothesis (BF₁₀ = 0.17). We also did not find that the ANX group was significantly slower at position 1 following a position 4 error compared to the HC group (between-groups rmANOVA; F(1,58) = 0.36, p = 0.55; η_p² = 0.0061; Figure 3B), although there was anecdotal (inconclusive) evidence (BF₁₀ = 0.67) for this null result. Therefore, we did not observe differences in the post error slowing effect between ANX and HC, which also supports the hypothesis that the observed deficit in initiation cost is related to a difference in abstract sequential control rather than more general cognitive control.

The last component of our first hypothesis was that symptom severity would correlate with sequence initiation deficits in OCD. Even though we did not observe the hypothesized sequence initiation deficits in OCD, we tested this hypothesis by correlating initiation cost and Y-BOCS scores as a measure of symptom severity in OCD. We did not observe a reliable correlation (Y-BOCS total: p = 0.40, r = −0.13; Y-BOCS compulsions subscale: p = 0.89, r = −0.02; Y-BOCS obsessions subscale: p = 0.08, r = 0.26). In light of the finding that ANX did have impaired initiation costs, we performed an exploratory analysis to correlate a measure of symptom severity in ANX, the APPQ measure, and initiation cost in that group. We did not observe a reliable correlation (APPQ total: p = 0.49, r = 0.-15; APPQ social anxiety disorder subscale: p = 0.24, r = 0.26; APPQ agoraphobia subscale: p = 0.88, r = 0.03; APPQ panic disorder subscale: p = 0.85, r = −0.04).

Our second hypothesis was that participants with OCD would exhibit a deficit in flexible cognitive control (measured with task switching) compared to the HC and ANX groups. To test this hypothesis, we compared RT and ER switch costs in OCD to HC and OCD to ANX groups. Our observations did not support the hypothesis. There were no significant differences for these group comparisons in switch costs in either RTs (Table 5; Figure 2B) or ER (Table 5; Figure 2D). For completeness and given the observed differences in initiation cost between the ANX and HC groups, we compared switch costs in these two groups as well and also found no differences in switch costs in RT (Table 5; Figure 2B) or ER (Table 5; Figure 2D). There was moderate evidence for the null hypotheses in these comparisons (RT: OCD vs. HC BF₁₀ = 0.19, OCD vs. ANX BF₁₀ = 0.2; ER: OCD vs. HC BF₁₀ = 0.15, OCD vs. ANX BF₁₀ = 0.2).

Table 5. Between groups rmANOVA for RT (s) and ER (%) switch cost.

Repeated measures ANOVAs were conducted between groups for switches – repeats in both reaction times (s) and error rates (%). Statistics are reported for group, trial type (switches or repeats), and group x trial type effects (factors). F statistics (d.f.s), P values, and effect sizes (η_p²) are reported for each factor.

		Reaction time			Error rate
Group	Factor	F Statistic (d.f.s)	P	η p 2	F Statistic (d.f.s)	P	η p 2
OCD vs. HC	Group	0.08 (87)	0.78	<0.001	0.06 (87)	0.81	<0.001
	Trial Type	19.89 (87)	<0.001	0.19	2.77 (87)	0.10	0.03
	Group x Trial Type	0.35 (2,87)	0.56	0.01	<0.001 (2,87)	0.98	<0.001
OCD vs. ANX	Group	3.33 (62)	0.07	0.05	0.06 (62)	0.81	<0.001
	Trial Type	18.71 (62)	<0.001	0.23	8.41 (62)	0.01	0.12
	Group x Trial Type	0.18 (2,62)	0.67	0.01	0.45 (2,62)	0.5	0.01
ANX vs. HC	Group	3.58 (66)	0.06	0.05	0.14 (66)	0.71	0.07
	Trial Type	11.22 (66)	0.01	0.15	5.12 (66)	0.03	0.01
	Group x Trial Type	0.01 (2,66)	0.93	<0.001	0.81 (2,66)	0.37	0.01

To test the hypothesis that symptom severity correlates with task switching deficits in OCD, we correlated RT and ER switch costs in this group with Y-BOCS scores. Perhaps unsurprisingly, due to the lack of task switching deficits in OCD, our hypothesis was not supported. We did not observe reliable correlations between Y-BOCS scores and RT switch costs (Y-BOCS total: p = 0.11, r = −0.25; Y-BOCS compulsions subscale: p = 0.25, r = −0.18; Y-BOCS obsessions subscale: p = 0.084, r = −0.27) or ER switch costs (Y-BOCS total: p = 0.42, r = −0.13; Y-BOCS compulsions subscale: p = 0.29, r = −0.17; Y-BOCS obsessions subscale: p = 0.69, r = −0.062). For completeness, as an exploratory analysis we tested if anxiety scores (APPQ total) correlated with RT and ER switch costs in the ANX groups. We did not observe a reliable correlation between APPQ scores and RT switch costs (p = 0.82, r = −0.05) or ER switch costs (p = 0.4, r = 0.19).

Discussion

This study investigated abstract sequential control and more general, flexible cognitive control in participants with OCD and anxiety disorders using abstract task sequences. Surprisingly, and in contrast to our hypotheses, we found that an indicator of sequential control (sequence initiation) was disrupted in the ANX but not the OCD group. Participants with ANX exhibited greater RTs and initiation costs compared to the OCD and HC groups. Further, we did not observe any differences across the groups in a measure of flexible cognitive control, task switching. Initiation costs did not correlate significantly with symptom severity in either clinical group, and switch costs did not correlate significantly with symptom severity in OCD or with anxiety scores in ANX. Together, these results are the first to show specific abstract sequential control differences in individuals with anxiety disorders and reveal new behavioral axes and potential neural differences to investigate that may dissociate OCD from ANX symptomatology. Specifically, sequential behavior in ANX may be more specific to deficits in abstraction and rely on rostrolateral PFC, while sequential behavior specific to OCD may more closely align with motor sequence deficits (not examined in this study), invoking motor related PFC regions and striatum.

The unexpected sequence initiation results in the OCD group inform our current understanding of OCD sequential behavior and neurobiological models that support it. We expected OCD participants to exhibit an abstract task sequence deficit based on previous studies showing impaired implicit sequence learning in this group (Goldman et al., 2008; Kathmann et al., 2005; Kelmendi et al., 2016). However, since we did not observe abstract task sequence deficits in OCD compared to HCs or the ANX group, sequential behavior in OCD may be better explained by dysfunction in circuitry responsible for implicit motor sequence learning rather than a common sequencing mechanism between implicit and explicit sequences. In one implicit sequence study, after performing implicit sequences, half the participants were provided with explicit knowledge of the sequential pattern. The OCD participants with this explicit instruction performed the task better than previously during the implicit portion, and better than matched controls (Soref et al., 2018). These and our current results suggest that OCD individuals may rely on brain structures that support explicit knowledge when exerting sequential control, and experience deficits only during implicit sequence processing. As outlined in neurobiological models, striatal regions important for implicit processing (Reiss et al., 2005) may be dysfunctional during sequential behavior in OCD while rostral PFC regions implicated in abstract sequential control remain relatively functional (Desrochers et al., 2015, 2019). Given this study was the first of its kind and had a limited sample size, further studies are needed to probe abstract task sequential control in OCD and the possible dissociation that exists between implicit motor and abstract task sequential control in this group. Such a dissociation may be valuable for diagnosis and suggests the striatum (instead of rostral PFC) may be a productive treatment target for maladaptive sequential behaviors observed in OCD.

The unexpected sequence initiation deficit observed in ANX suggests a possible dissociation between OCD and anxiety disorders along the axis of sequential control. Since our initial predictions in this task were based on implicit sequence literature, our finding was unexpected as implicit sequence learning has been shown to be intact in people with anxiety disorders (including social anxiety disorder, agoraphobia, and panic disorder) (Goldman et al., 2008) compared to those with OCD. However, a phenomenon that affects behavior in abstract sequential control but not during implicit sequence tasks is serial attention, in which behavior on an upcoming sequence step is affected by allocation of attention on previous steps (Desrochers et al., 2022). It has been shown that people with anxiety disorders exhibit deficits in attentional control (Rueppel et al., 2022; Yu et al., 2018), and set-shifting (Kertz, Belden, Tillman, & Luby, 2016), the ability to unconsciously shift attention from one task to another. Speculatively, deficits in attention and shifting could lead to slower RTs in abstract task sequences, as it requires both. If sequence control and task control require at least partially shared resources, as suggested by the ability for each to influence the other (Trach et al., 2021), then the increased RT deficits in ANX specifically at the first position could reflect a multiplicative effect of needing to engage additional, hierarchical control and attentional processes necessary for sequential control (Schneider & Logan, 2006). Abstract task sequential control deficits therefore may contribute to the literature on the role of attention in ANX symptom manifestation and motivate future studies. Additionally, these results suggest a tool to dissociate OCD from other non-OCD anxiety disorders may be abstract sequential task performance, which may inform clinical theories about these disorders. Specifically, clinical theories of anxiety may incorporate abstract task sequential behavior deficits as a feature of anxiety disorders, potentially providing more precise clinical diagnostic criteria for this group of disorders compared to OCD.

This study was limited primarily because of the natural heterogeneity in diagnostic classification, resulting in difficulty isolating diagnostic groups due to comorbidities within participants, which resulted in small group sizes. Heterogeneity was perhaps greatest in the ANX group, which contained the smallest sample size (n = 22) and was composed of participants with a primary diagnosis of any of the three assessed anxiety disorders. Further, recruitment of participants was optimized for a larger study with different primary research questions and not for the questions posed in this current study. For example, in correlating behavior with clinical measures in ANX, the APPQ was our primary measure. It may have been useful to obtain more dimensional clinical severity data using a transdiagnostic measure such as the Overall Anxiety and Stress Impairment Scale (Norman, Cissell, Means-Christensen, & Stein, 2006). Both factors may contribute to the marginally significant effect in initiation costs observed between the OCD and ANX groups, and the medium effect sizes we observed in some results. Though these conditions were not ideal, the finding of an initiation cost deficit that possibly dissociates anxiety disorders from OCD is notable. However, as this study was the first to investigate abstract sequential task control in OCD and ANX, similar studies should be conducted before drawing strong conclusions about these dissociations. Nonetheless, this research highlights the utility of studies using large sample sizes across an array of measures and calls for future studies investigating abstract task sequential control in both groups.

Our results highlight the need for more studies of task switching in OCD during complex tasks that engage multiple cognitive demands that extend through time, such as the present behavioral paradigm, because such paradigms may reveal compensatory mechanisms and more closely approximate behavior in everyday life. The majority of previous task switching studies in OCD observed deficits in this process compared to HCs, but task switching was studied in relative isolation (Gu et al., 2008; Han et al., 2011; Liu et al., 2023; Meiran et al., 2011; Remijnse et al., 2013). One, relatively uncommon, study that examined task switching in conjunction with a secondary task (“inhibition-of-switching”) did not find switch cost deficits in OCD compared to HCs (Demeter, Harsányi, Csigó, & Racsmány, 2017). Therefore, it is possible that the secondary task, increased multiple cognitive demands (e.g., cognitive control, attention, working memory), and/or being nested within higher-order sequential structure in the current experiment, engages additional, potentially compensatory resources. Future studies could also be designed to elicit and dissociate more error types (e.g., perception, slip of action, wrong judgement) to aid in determining how information is maintained throughout and may contribute to task demands. Further study is necessary to disentangle these possibilities, to determine if and what kinds of additional tasks may obscure task switching deficits in OCD and the nature of specific errors made to better understand how sequential behavior unfolds in daily life.

Our results suggest sequential behavior deficits in OCD may occur primarily in implicit motor rather than abstract task sequence processing, implicating specific neural circuitry and potentially guiding future treatments. Specifically, since implicit motor sequence deficits have been observed in OCD (Goldman et al., 2008) and this process relies on the striatum (Reiss et al., 2005), the striatal component of the CSTC model may be more heavily implicated than prefrontal regions in supporting sequential symptomatology in OCD. In the future, this work could contribute to more individualized treatments for OCD patients who exhibit sequential behavior deficits. For example, clinicians may favor modulatory treatments, such as repetitive transcranial magnetic stimulation to the dorsomedial (Dunlop et al., 2016) or dorsolateral (Caparelli et al., 2022) PFC, that indirectly impact the striatum by targeting connected cortical regions when treating patients presenting with dysfunctional sequential behavior. Overall, our work advances the current understand of the role of sequential behavior in OCD, informing current neurobiological models and potentially impacting future treatments.

Although no reliable behavioral differences were observed in OCD in the present sequential paradigm, the underlying neural circuitry may differ between OCD and HCs and abstract task sequence performance may be accomplished differently in these two groups. Similar behavior between groups, particularly in clinical groups, may be caused by different neural mechanisms (Huys, Maia, & Frank, 2016). Further, neural circuits implicated in OCD, especially in prefrontal and striatal regions, overlap with brain regions involved in abstract task sequence processing (Alexander, DeLong, & Strick, 1986; Desrochers et al., 2022, 2015, 2019; Greenberg et al., 1997; Harrison et al., 2009; Page et al., 2009; Roth et al., 2007). This overlap suggests that underlying neural activity may be disrupted in OCD and necessitates the involvement of alternative neural circuitry to accomplish abstract task sequence performance that is comparable to HCs. Therefore, although no behavioral differences between OCD and HC were observed in our task, an open question remains as to whether this group exhibits a distinct neural mechanism to achieve abstract task sequential control compared to HCs.

Though we did not originally hypothesize that ANX would exhibit sequential control deficits, previous studies illustrate the potential overlap in underlying neural circuitry between anxiety disorders and abstract task sequential control and suggest a potential avenue for future investigation. In anxiety disorders, neuroimaging studies point to circuitry involving prefrontal and subcortical regions that underlie pathology (Berkowitz, Coplan, Reddy, & Gorman, 2007). Although compulsions in ANX are thought to be supported by neural circuitry associated with a fear response (Duval et al., 2015; Goodwin, 2015), hypoactivation of the PFC and hyperactivation of the amygdala can co-occur with such behavior in anxiety disorders (Shin & Liberzon, 2010), implicating the dysfunction of PFC regions in these disorders. Further, prefrontal cortex dysfunction has been observed in participants with generalized anxiety disorder during an emotion dysregulation task (Mochcovitch, da Rocha Freire, Garcia, & Nardi, 2014) and participants with high anxiety exhibit impaired cognitive control (J. Wood, Mathews, & Dalgleish, 2001), which involves frontal cortex. These prefrontal circuits implicated in anxiety disorders overlap with those shown to be necessary for abstract task sequential control in the rostral lateral prefrontal cortex (Desrochers et al., 2015, 2019). Shared symptoms between OCD and anxiety disorders, such as repetitive behaviors, may therefore share similar underlying prefrontal but different subcortical neural circuitry. Additional studies are needed to directly investigate this potential link.

Using abstract task sequences, we showed that ANX participants exhibit an initiation deficit in abstract task sequential control compared to HCs (with a marginal effect compared to OCD). The findings from this study provide a novel framework under which to interpret both OCD and ANX symptomatology, suggesting OCD sequential behavior may be explained more clearly by implicit motor sequence deficits and showing a dissociation between OCD and ANX during abstract task sequences. Results from this study enhance understanding of the unique neuropathology of OCD and anxiety disorders and will provide the foundation for future studies investigating the neural bases of abstract task sequential control in clinical groups.